Recent Advances in Parkinson’s Disease Recognition from EEG Using Attention-Based Sparse Graph Convolutional Neural Networks: A Systematic Review

Abstract

Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by motor and cognitive impairments, making early diagnosis crucial for effective intervention. Electroencephalography (EEG) has emerged as a promising non-invasive modality for detecting PD-related neural abnormalities. Recent advances in artificial intelligence, particularly deep learning, have significantly enhanced EEG-based PD recognition. Among these, attention-based sparse graph convolutional neural networks (ASGCNN) have demonstrated superior capability by modeling functional brain connectivity and selectively focusing on discriminative EEG channels. This systematic review presents recent developments in EEG-based PD detection, focusing on graph neural networks (GNN), attention mechanisms, and hybrid deep learning architectures. Graph-based models effectively capture spatial relationships among EEG channels, while attention mechanisms improve interpretability and feature selection. The incorporation of sparsity constraints reduces redundancy and enhances computational efficiency. Recent studies report classification accuracies exceeding 90–97%, outperforming conventional machine learning and CNN-based approaches.  The review also highlights key challenges such as data heterogeneity, lack of interpretability, and computational complexity. Emerging trends include explainable AI, multimodal fusion, and wearable EEG systems. Overall, ASGCNN-based frameworks represent a promising direction for accurate, interpretable, and scalable PD diagnosis.